Hello, I have been using the Hailo Dataflow Compiler to compile a custom YoloV11 model. I followed the guide Guide to using the DFC to convert a modified YoloV11 on Google Colab - Guides - Hailo Community.
The model compilation was successful; however, when I use the model in my Python code for inference on Hailo, specifically this line:
results = hailo.run(frame)
it returns “results” as a Python dictionary with 6 elements instead of a list, like precompiled YOLO models in the Model Zoo usually return. Upon analysis, it seems that the postprocessing did not occur correctly.
I tried compiling my model again but this time using CLI commands instead of following the steps from the guide, and it worked , results = hailo.run(frame) returned a list as expected.
What could be the reason why following the guide steps results in incorrect postprocessing, whereas using CLI commands works as intended?
Hi @dgarrido,
Based on your description, the different outputs you’re seeing are likely due to HailoRT post-processing being included in one HEF and not the other. Without seeing the actual outputs, I can’t specify which is which, but you can check your YAML configurations to confirm this.
Let me know if you’d like me to help review your configurations!
Best regards,
Omria